Use the provided Dockerfile to build a Docker image containing the compilation environment.
In short, run docker build -t fgpu_compiler . inside this directory.
The compiler binaries will then be available inside the container in the /llvm/llvm-3.7.1.build/bin folder.
To compile the kernels, you have to bring the kernels directory into the Docker container. An example usage is provided below:
- Do
docker run -v $(pwd):/llvm/kernels -it fgpu_compilerfrom inside the kernels directory. This will create a directory/llvm/kernelsinside the container and mount the current directory over there, and then run the container into interactive mode (i.e., you get a shell inside the container). - Inside the container,
cdinto the kernels folder and run thecompile.shscript, following its instructions accordingly.